Simulating the cross-linguistic pattern of optional infinitive errors in children’s declaratives and wh- questions [forthcoming]

One of the most striking features of children’s early multi-word speech is their tendency to produce non-finite verb forms in contexts in which a finite verb form is required (Optional Infinitive [OI] errors, Wexler, 1994). MOSAIC is a computational model of language learning that simulates developmental changes in the rate of OI errors across several different languages by learning compound finite constructions from the right edge of the utterance (Freudenthal, Pine & Gobet, 2006a; 2009; Freudenthal, Pine, Aguado-Orea & Gobet, 2007). However, MOSAIC currently only simulates the pattern of OI errors in declaratives, and there are important differences in the cross-linguistic patterning of OI errors in declaratives and Wh- questions. In the present study, we describe a new version of MOSAIC that learns from both the right and left edges of the utterance. Our simulations demonstrate that this new version of the model is able to capture the cross-linguistic patterning of OI errors in declaratives in English, Dutch, German and Spanish by learning from declarative input, and the cross-linguistic patterning of OI errors in Wh- questions in English, German and Spanish by learning from interrogative input. These results show that MOSAIC is able to provide an integrated account of the cross-linguistic patterning of OI errors in declaratives and Wh- questions, and provide further support for the view, instantiated in MOSAIC, that OI errors are compound-finite utterances with missing modals or auxiliaries